Mobile App Quality Assessment: Leveraging LLM-Driven Data Augmentation for Enhanced User Reviews Analysis
Bassam Alsanousi, Stephanie Ann Ludi, Hyunsook Do · 2025
Mobile applications have been developing at an extraordinary pace, with high-quality apps being required for meeting user expectations while remaining competitive in the market. In this research, we propose an approach to assess mobile app quality by leveraging Large Language Models (LLMs) for classifying user reviews according to the ISO/IEC 25010 quality standard. To enhance the performance of the classification model in cases of limited or imbalanced datasets, we utilized OpenAI's GPT-4o model for performing data augmentation (DA) on labeled user reviews according to the ISO/IEC quality standard. Then, we compared our approach against two different fine-tuned advanced LLMs (GPT-4o and Gemini 1.0 Pro) in two scenarios: without and with DA (non-DA and DA). Our findings have shown a clear boost in performance when using DA compared to non-DA, GPT-4o, and Gemini 1.0 Pro. Finally, we incorporated Local Interpretable Model-agnostic Explanations (LIME) to interpret the output predictions of our proposed approach.